Include hasWarn and textures in the shader rebuild key

[kmatzen]

Fixes #55. Stacked on #57 — base is fix-gpu-resource-leak, so this shows only its own diff and will retarget to main when #57 merges.

The rebuild decision compared only glsl and paramCount:

if (sdf.glsl !== this.lastGlsl || sdf.paramCount !== this.lastParamCount) {

but rebuild() bakes two more inputs into the material — hasWarn selects the fragment source via buildFrag(sdf.glsl, sdf.paramCount, sdf.hasWarn), and each texture becomes a uniform. An evaluation producing identical GLSL and param count but a different hasWarn therefore never reached the GPU, leaving the thin-wall highlight showing the previous evaluation's state.

Correction to the issue as I filed it

The textures half is latent, not live. codegen.ts declares textures (:6), resets it (:324), and returns a copy (:352), but nothing ever pushes to it:

$ grep -nE 'textures\s*(=|\.push|\.splice|\[)' src/worker/sdf/codegen.ts
324:  textures = [];

Nothing else in src/worker/sdf/ mentions textures. So sdf.textures is always [] at runtime and a texture-only change cannot currently produce a stale render. I over-claimed a user-visible symptom that can't presently occur, and have said so on the issue rather than quietly narrowing the fix.

Only hasWarn is a live bug. The texture key is included anyway — it costs nothing while the array is empty, and it means the key is already correct whenever textures get wired up, rather than being a trap waiting for that change. It hashes content (FNV-1a) rather than just dimensions, so a content edit at unchanged width/height still invalidates.

Verification

check	result
tsc --noEmit	exit 0
SdfMesh tests, this branch	7/7 pass
new tests, base branch	2 fail, 1 pass

The test that passes on both asserts an identical display does not rebuild. That guards the opposite direction: each rebuild recompiles a shader and (after #57) disposes the previous one, so an over-eager key would trade a stale-render bug for a churn bug.

🤖 Generated with Claude Code

[kmatzen]

Can't reopen this — GitHub refuses (HTTP 422) because the head branch was force-pushed (rebased onto main) after the PR auto-closed. Re-filed as #67 with the same commit and body. Closing the loop here.